bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Acquaviva et al., 30 Jan, 2019 – bioRxiv preprint v1

Ensuring meiotic DNA break formation in the mouse pseudoautosomal region

Laurent Acquaviva1*, Michiel Boekhout1†, Mehmet E. Karasu1,2, Kevin Brick3, Florencia Pratto3, Megan van Overbeek1‡, Liisa Kauppi1§, R. Daniel Camerini-Otero3, Maria Jasin2,4* and Scott Keeney1,2,5*

1 Molecular Biology Program, Memorial Sloan Kettering Cancer Center, New York, NY, USA. 2 Louis V. Gerstner, Jr., Graduate School of Biomedical Sciences, Memorial Sloan Kettering Cancer Center, New York, NY, USA. 3 Genetics & Biochemistry Branch, NIDDK, NIH, Bethesda, MD, USA. 4 Developmental Biology Program, Memorial Sloan Kettering Cancer Center, New York, NY, USA. 5 Howard Hughes Medical Institute, Memorial Sloan Kettering Cancer Center, New York, NY, USA.

* Correspondence to [email protected], [email protected] or [email protected]

Current addresses: †UMC Utrecht, Oncode Insitute, Utrecht University, Utrecht, Netherlands; ‡Caribou Biosciences, Inc., Berkeley, CA, USA; §Faculty of Medicine, University of Helsinki, Helsinki, Finland.

Sex in males share only a diminutive homologous Results segment, the pseudoautosomal region (PAR), wherein meiotic double-strand breaks (DSBs), pairing, and crossing over must occur A distinctive PAR ultrastructure rich in pro-DSB factors for correct segregation. How cells ensure PAR recombination is We applied a cytogenetic approach to investigate the mouse PAR unknown. Here we delineate cis- and trans-acting factors that control structure in the C57BL/6J strain (B6). In most of the genome, axes PAR ultrastructure and make the PAR the hottest area of DSB elongate and DSBs begin to form during leptonema, then homologous formation in the male mouse genome. Prior to DSB formation, PAR chromosomes pair and axes are juxtaposed by the transverse filament axes elongate, sister chromatids separate, and DSB- SYCP1 (forming the synaptonemal complex, or SC) during promoting factors hyperaccumulate. These phenomena are linked to zygonema. and recombination are completed during mo-2 minisatellite arrays and require ANKRD31 protein. We pachynema, then the SC disassembles during diplonema with homologs propose that the repetitive PAR sequence confers unique chromatin remaining attached at sites of crossovers (chiasmata) until anaphase I (1, and higher order structures crucial for DSB formation, X–Y pairing, 27). X and Y usually pair late, with PARs paired in less than 20% of and recombination. Our findings establish a mechanistic paradigm spermatocytes at late zygonema when nearly all are paired (7, of mammalian segregation during spermatogenesis. 8). At this stage, we found by conventional immunofluorescence microscopy that unsynapsed PAR axial elements (SYCP2/3 staining) Introduction appeared thickened relative to other unsynapsed axes and had bright HORMAD1/2 staining (Fig. 1A and fig. S1A,B) (28). Meiotic recombination forms connections between homologous We found that the PAR was highly enriched for DSB-promoting chromosomes that ensure accurate segregation (1). In many species, factors REC114, MEI4, MEI1, and IHO1 [ essential for genome- every chromosome must recombine, so a crucial challenge is to ensure wide DSB formation (22-24, 29)] as well as ANKRD31, a REC114 that every chromosome pair acquires at least one SPO11-generated DSB partner essential for PAR DSB formation (30, 31). All five proteins to initiate recombination (2). (hereafter RMMAI for simplicity) colocalized in several bright, irregular This challenge is especially acute in most male placental mammals “blobs” for most of prophase I (Fig. 1A–C and fig. S1C). Two blobs for sex chromosomes (X and Y), on which a DSB can only support were on the X and Y PARs as judged by chromosome morphology at late recombination if it occurs in the tiny PAR (3-7). The PAR in laboratory zygonema (Fig. 1A) and particularly bright fluorescence in situ mice is the shortest thus far mapped in mammals, at ~700 kb (5, 6). Since hybridization (FISH) with a probe for the PAR boundary (PARb) (Fig. only one DSB is formed per ten megabases on average in the mouse, the 1C). Other blobs highlighted the distal ends of specific autosomes (Fig. PAR would risk frequent recombination failure if it behaved like a typical 1C), which we revisit below. Undefined blobs are also apparent in autosomal segment (7, 8). However, the PAR is not typical, having published micrographs but were not explored further (21-24). Consistent disproportionately frequent DSB formation and recombination (4, 7, 9, with and extending other studies (21-24, 30, 31), all five proteins also 10). The mechanisms promoting such frequent DSBs are not known in colocalized in numerous small foci along unsynapsed chromosome axes any species. (Fig. 1B and fig. S1C), but PAR staining was much brighter. ANKRD31, Higher order chromosome structure plays an important role in MEI1 and REC114 enrichment on the PAR was already detectable in pre- meiotic recombination. DSBs arise concomitantly with development of leptotene cells during premeiotic S phase (fig. S1D), as shown for MEI4 linear axial structures that anchor arrays of chromatin loops within which and IHO1 (21, 23). DSBs occur (1, 11-13). The axis begins to form between sister Structured illumination microscopy (SIM) resolved the thickened chromatids during pre-meiotic replication (pre-leptonema) and includes PAR axes as two strands of axial core (Fig. 1D and fig. S2A,B) that were SYCP2 and SYCP3 (14, 15), cohesin complexes (16), and HORMA heavily decorated along their lengths with RMMAI proteins (Fig. 1E). domain proteins (HORMAD1 and HORMAD2) (17-20). Axes are also At the zygotene–pachytene transition, X and Y pair and initiate synapsis, assembly sites for IHO1, MEI4, and REC114 complexes (13, 21-24), then SC spreads bidirectionally—homologously in the PAR and non- whose functions as essential promoters of SPO11 activity are homologously between the non-PAR axes, which definitively marks incompletely understood (25, 26). early pachynema in mice (28, 32, 33) (Fig. 1D). Separation of PAR axes We previously showed that PAR chromatin is organized into short was readily observed by SIM in late zygonema before X and Y pairing loops on a long axis, (7). However, only a low-resolution view of PAR and synapsis, remained apparent after synapsis, then disappeared during structure was available and the cis- and trans-acting factors controlling early pachynema (Fig. 1D). A multi-core structure was also seen in PAR structure and DSB formation have remained largely unknown. earlier electron microscopy studies, but was staged incorrectly as occurring at mid-to-late pachynema (34) (fig. S2C,D).

1 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Sex chromosome recombination

In principle, the sister chromatid axes could be split apart (Fig. 1F) and the association of compact PARb chromatin with a long axis. or the PAR could be folded back on itself in a crozier configuration (fig. S2E). However, a crozier was ruled out by using SIM to define the path RMMAI proteins accumulate at mo-2 minisatellites of the PAR: the binding protein TRF1 (35) decorated only the What are the cis-acting determinants of this PAR behavior? We tip of the chromosome (Fig. 1G) and the PARb probe yielded FISH deduced that DNA sequences shared between the PAR and certain signals symmetrically arrayed at similar positions on both axial cores autosomes might be responsible for recruiting RMMAI proteins because (fig. S2F). We conclude that each axial core is one of the sister the blobs consistently decorated the distal tips of specific autosomes that chromatids, with a “bubble” opened from near the PAR boundary almost also hybridized, albeit more weakly, to the PARb FISH probe (Fig. 1C). to the telomere (Fig. 1F). Axis splitting and REC114 enrichment Self-aligning the PARb probe sequence illustrated its highly occurred in the absence of SPO11 (Fig. 1H), thus neither property of the repetitive nature, including a central ~20-kb tandem array of a PAR is provoked by DSB formation. minisatellite called mo-2, with a 31-bp GC-rich repeat (36, 37) (Fig. 3A). These findings establish that the X and Y PARs adopt an elaborate BLAST searches detected additional mo-2 clusters at the non- axial structure that forms prior to homologous pairing and synapsis and centromeric ends of chr9 and chr13 in the mm10 genome assembly and disappears during early pachynema. It normally occurs at or before the chr4 from the Celera assembly (Fig. 3A), matching the reported time when most cells make PAR DSBs (which is in late zygonema or distribution (36, 37). Although these autosomal clusters are only 2 to 8 around the zygotene–pachytene transition (7)) but also forms in the kb in the assemblies, their copy number is likely underrepresented absence of DSBs and is preceded by accumulation of high levels of because they lie adjacent to gaps. RMMAI proteins, which starts in pre-leptonema. To test if RMMAI blobs correspond to mo-2 arrays, we used an oligonucleotide FISH probe with the mo-2 consensus sequence. This Dynamic remodeling of the PAR loop–axis structure probe gave a compact signal along the PAR axes that was similar to the To investigate temporal patterns of axis differentiation, RMMAI RMMAI blob pattern in shape and proportional intensity (Fig. 3B and composition, and chromatin loop configuration, we used SIM to compare fig. S4A,B). We confirmed the identity of the autosomes with localization of ANKRD31, SYCP3, and the meiotic cohesin subunit chromosome-specific probes (fig. S4C). REC8 from pre-leptonema to mid-pachynema (Fig. 2A). In parallel, we These findings led us to hypothesize that mo-2 arrays might be a cis- used conventional microscopy and immunoFISH to measure axis lengths acting determinant of RMMAI recruitment. If so, we reasoned that (distance from the PARb probe to the end of the SYCP3-labeled axis), reducing mo-2 copy number might reduce the amount of RMMAI REC114 spreading along the axis, and loop sizes defined as axis- proteins. To test this prediction, we took advantage of a natural orthogonal extension of the PARb FISH signal (Fig. 2B and fig. S3). experiment afforded by mouse strain diversity, namely, the fact that the This analysis delineated a dynamic, large-scale reconfiguration of PAR Mus musculus molossinus subspecies has substantially lower mo-2 copy loop–axis structure as prophase I proceeds. number on the PAR and autosomes (37). The wild-derived M. m. At pre-leptonema, ANKRD31 blobs had a closely juxtaposed REC8 molossinus strain MSM/MsJ (hereafter MSM) showed less hybridization focus (Fig. 2A: chromosome a). In leptonema and early zygonema, signal with the mo-2 FISH probe as expected, but importantly also gave ANKRD31 and REC114 signals stretched along the length of the lower intensity of REC114 immunostaining in blobs compared to B6 (fig. presumptive PAR axes, while REC8 was restricted to the proximal and S4D). distal borders (Fig. 2A:b–e and 2Bi). The SYCP3-defined axial element To quantify this difference and to avoid confounding effects from was already long as soon as it was detectable (0.72 µm on average for the strain differences in trans-acting factors, we examined spermatocyte mouse in Fig. 2B) and the PARb FISH signal was compact (0.53 µm) spreads from reciprocal F1 hybrid offspring from MSM and B6 parents (Fig. 2Bi), confirming and extending our previous observation of long (Fig. 3C and fig. S4E). As we predicted, less ANKRD31 accumulated PAR axes and short loops at this stage relative to autosomal loci (7). on MSM-derived PARs. That is, the YMSM PAR had 8-fold less At late zygonema, the PAR axis had lengthened still further (0.93 ANKRD31 than the XB6 PAR in F1 hybrids from B6 mothers and MSM µm), while the PARb signal remained compact (Fig. 2Bii). It was during fathers (Fig. 3Ci), and the XMSM PAR had 6.5-fold less than the YB6 PAR this stage that the PAR split into separate axial cores, each carrying high in F1 hybrids from the reciprocal cross (Fig. 3Cii). These relative levels of RMMAI proteins (Fig. 2A:f–h). The split corresponded to a ANKRD31 intensities closely matched relative intensities of mo-2 FISH REC8-poor zone bounded proximally and distally by focally enriched signals. Despite these quantitative differences, MSM PARs clearly meet REC8 (Fig. 1F and 2A:f–h). the minimal requirements to support sex chromosome pairing with As cells transitioned into early pachynema and the X and Y PARs efficiency and timing similar to B6 (fig. S4F), not suprisingly since M. synapsed (Fig. 2A:i–m), the PAR axes began to shorten slightly (0.86 m. molossinus is a fully fertile subspecies. Interestingly, however, µm) while the PARb signal expanded (0.85 µm) (Fig. 2Biii). Meanwhile, although staining for the ssDNA binding protein RPA was still present the elongated ANKRD31 signals progressively decreased in intensity, on MSM PARs it was nonetheless lower in intensity (Fig. 3C). We revisit collapsed along with the shortening axes, and separated from the axis this observation below. while remaining in its vicinity (Fig. 2A:l–m). By mid-pachynema, PAR axes collapsed still further, to about half their zygotene length (0.47 µm) Both RMMAI recruitment and PAR loop–axis remodeling require and the PARb chromatin expanded to more than twice the zygotene ANKRD31 and MEI4 measurement (1.3 µm) (Fig. 2Biv). ANKRD31 and REC114 enrichment To identify trans-acting factors important for PAR behavior, we largely disappeared during this stage, leaving behind a bright bolus of tested eliminating RMMAI components or axis proteins. We REC8 on the short remaining PAR axis (Fig. 2A:n–o and 2Biv). counted the number of axial RMMAI foci genome wide and assessed mo- To sum up, the PAR already has a long axis and compact chromatin 2-associated blobs in leptotene/early-zygotene spermatocytes (Fig. 4A loops as soon as the axis forms at the beginning of prophase, and the axis and fig. S5A), and examined PAR structure in late zygonema by lengthens and sister axes split apart in late zygonema before X and Y immuno-FISH for REC114 and mo-2 by conventional microscopy (Fig. pairing. After synapsis, the axes shorten and the chromatin loops 4B and fig. S5B) and SIM (Fig. 4C). decompact. RMMAI proteins are enriched before axis formation and they In Rec8–/– (38), RMMAI foci formed at or slightly above normal spread along the axis as it forms and lengthens, then dissociate numbers along unsynapsed axes [as shown for MEI4 (21)] and RMMAI concomitant with axis collapse in pachynema. This analysis establishes blobs formed on mo-2 regions (Fig. 4A and fig. S5A). Moreover, axis spatial and temporal correlations between presence of RMMAI proteins elongation, splitting of sister axes, and formation of short loops (i.e.,

2 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Acquaviva et al., 30 Jan, 2019 – bioRxiv preprint v1 compact mo-2 and REC114 signals) all occurred normally (Fig. 4B,C These predictions were met. and fig. S5B). However, the PAR axes were separated at the telomeric To evaluate DSB formation, we quantified the number of axial end and the FISH signals for each sister chromatid in the distal PAR were RPA2 foci as a proxy for global DSB numbers, and assessed the widely separated (Fig. 4C and fig. S5C). We conclude that REC8 proportion of mo-2 FISH signals that contained an RPA2 focus (Fig. 5C– maintains cohesion in the distal PAR but is dispensable for RMMAI E and fig. S7A). Wild-type spermatocytes had a mean of 165 RPA2 foci assembly and formation of the specialized PAR structure. per cell in zygonema, which declined to 96 foci per cell in pachynema as We tested the importance of RMMAI proteins using Mei4–/– (22) DSB repair progressed (Fig. 5D). In zygonema, RPA2 foci overlapped and Ankrd31–/– (30) mutants. In the absence of MEI4, IHO1 foci were not on average 35% of each cell’s mo-2 regions, increasing to 70% at decreased (Fig. 4A and fig. S5A). Instead, they were slightly increased, pachynema (Fig. 5E), at which time 95% of spermatocytes had possibly reflecting stabilization if DSBs are not formed (23). In contrast, accomplished homologous X–Y pairing as measured by PAR FISH (Fig. REC114 and ANKRD31 were virtually absent from chromosome axes 5F). and did not form blobs on mo-2 regions. In Ankrd31–/– mice, MEI4 and Ankrd31–/– mutants had a stark reduction in the percentage of mo-2 REC114 foci still formed, but fewer and of lower intensity (Fig. 4A and FISH signals containing an RPA2 focus (Fig. 5E) even though global fig. S5A,D,E) (30). IHO1 foci formed in high numbers (as in Mei4–/–). RPA2 focus numbers were only modestly different from wild type at the Strikingly, however, RMMAI proteins did not accumulate detectably in stages assayed (Fig. 5D). [Ankrd31–/– mutants form substantially fewer mo-2-associated blobs (Fig. 4A and fig. S5A). Moreover, the normal RPA2 foci and other cytological DSB markers at leptonema and early PAR ultrastructure was absent in both Mei4–/– and Ankrd31–/– mutants: zygonema, but not from mid-zygonema on (30, 31).] As a consequence, axes were short and showed no sign of splitting, and the mo-2 FISH the X and Y chromosomes were paired in only 6% of mid-pachytene signal was highly extended (Fig. 4B,C and fig. S5B). MEI4 is thus spermatocytes (Fig. 5F), whereas most Ankrd31–/– cells pair and synapse essential for assembly of RMMAI foci genome wide, while ANKRD31 all autosomes (30, 31). contributes genome-wide but is more specifically essential for high-level To analyze mice with Rec8 and Hormad1 mutations, we used Syce1– enrichment at mo-2 regions. These results agree with findings of Tóth /– mutants as a point of comparison because they show similar meiotic and colleagues, who further observed that REC114 but not IHO1 is progression defects without defects in recruitment of RMMAI proteins. essential for formation of blobs of the other RMMAI proteins (31). SYCE1 is a component of the central element of the SC (39). Rec8 HORMAD1 is important for formation of MEI4 and IHO1 foci deficiency did not reduce RPA2 focus formation relative to Syce1–/–, genome wide (21, 23). In agreement, we found fewer foci of these and either globally on chromosome axes or specifically on mo-2 regions (Fig. other RMMAI proteins in the Hormad1–/– mutant (20) (Fig. 4A and 5D,E). Despite this, X–Y pairing was substantially reduced in pachytene- S5A). Remarkably, however, HORMAD1 was dispensable for RMMAI like spermatocytes (Fig. 5F), likely reflecting that REC8 promotes blobs (Fig. 4A and fig. S5A) and for PAR axial and chromatin structures interhomolog recombination in many species (40). (Fig. 4B,C and fig. S5B). As HORMAD1 is dispensable for RMMAI recruitment to mo-2 Several conclusions emerge. First, PAR RMMAI blobs share regions and for PAR ultrastructure, we predicted that it would also be genetic requirements with autosomal mo-2 blobs for their formation: dispensable for mo-2-associated DSBs. Indeed, Hormad1–/– Mei4 and Ankrd31 are essential but Rec8 and Hormad1 are dispensable. spermatocytes had comparable or higher frequencies of RPA2 foci Second, these requirements are distinct from those for formation of the overlapping mo-2 regions (Fig. 5E and fig. S7B) and of X–Y pairing as smaller, more widely dispersed RMMAI foci, i.e., Hormad1 is important the Syce1–/– control (Fig. 5F). The high frequency of mo-2-associated and Mei4 even more so, but Ankrd31 contributes only partially. Finally, RPA2 foci was striking given the substantial global reduction in RPA2 we establish a functional correlation between the ability to recruit high foci (Fig. 5D) and DSBs (19). levels of RMMAI to the PAR and the ability of the PAR to undergo its Collectively, these findings establish a tight correlation between normal structural differentiation. RMMAI recruitment (which itself correlates with axis remodeling) and high-frequency DSB formation. Further strengthening this correlation, PAR(-like) axis remodeling is tied to mo-2 presence and copy number we noted above that MSM-derived PARs display lower RPA2 staining If mo-2 arrays are cis-acting determinants of high-level RMMAI intensity (Fig. 3C). Although this result could mean that each DSB in an recruitment, and if RMMAI assemblies in turn govern PAR structural MSM PAR forms less ssDNA or binds less RPA2 for other reasons, we dynamics, then we would expect that autosomal mo-2 regions should also think it more likely that the lower RPA2 intensity reflects a lesser be able to form PAR-like structures. Indeed, we observed axis splitting tendency to make multiple DSBs. Indeed, multiple RPA2 foci could often at the distal end of chr9 (Fig. 5A and fig. S6A), which has the largest be resolved by SIM on PARs of B6 spermatocytes (fig. S7C), consistent autosomal mo-2 array. Moreover, REC114 immunofluorescence and mo- with prior reports of double PAR crossovers (4). Furthermore, examples 2 FISH signals showed the PAR-like pattern of extended axes and where multiple RPA2 foci were apparent on as-yet unsynapsed Y compact chromatin that was fully dependent on Ankrd31 (Fig. 5B). Thus, chromosomes were less frequent in MSM spermatocytes (fig. S7D). mo-2 arrays (and/or mo-2-linked elements) are apparently sufficient for To test more directly whether autosomal mo-2 regions experience both RMMAI recruitment and axis remodeling. PAR-like DSB formation, we used maps of ssDNA bound by the strand- We also examined PARs from M. m. molossinus. In keeping with exchange protein DMC1 (ssDNA sequencing, or SSDS) (9, 41, 42). reduced mo-2 and ANKRD31 (Fig. 3C), we observed substantially less Separate studies showed that ANKRD31 is critical for high-level DSB axis remodeling for MSM-derived PARs in the reciprocal F1 hybrid formation in the PAR (30, 31). We used our SSDS maps from whole- strains from B6 and MSM parents (fig. S6B). These findings reinforce testis samples of juvenile mice (12 days post partum (30)) and other the correlation between mo-2 presence, RMMAI levels, and PAR SSDS maps to test for PAR-like (i.e., ANKRD31-dependent) DSB signal ultrastructure. at autosomal mo-2 regions. Because of large gaps in the chromosome assemblies, neither the Determinants of PAR DSB formation in spermatocytes PAR nor the autosomal mo-2 regions can be fully assessed by SSDS or We hypothesized that RMMAI recruitment (and possibly also the other current deep-sequencing methods. Nonetheless, the region resulting axis remodeling) creates an environment conducive to high- encompassing the mo-2 cluster on chr9 displayed accumulation of SSDS level DSB formation in spermatocytes. This idea predicts that mutations reads that was substantially reduced in the Ankrd31–/– mutant (Fig. 5G should affect (or not) all of these processes coordinately and that and fig. S8A). A modest ANKRD31-dependent peak was also observed autosomal mo-2 regions should experience PAR-like DSB elevation. near the mo-2 cluster on chr13 (fig. S8B). We did not assess chr4 because

3 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Sex chromosome recombination available assemblies are too incomplete. The function of PAR axis splitting remains unclear. Sister Another hallmark of PAR DSB formation is that it is substantially chromatids develop individualized axes later in prophase (diplonema or independent of the histone methyltransferase PRDM9 (9). Similarly, diakinesis) in many species (1), and expansion microscopy of Drosophila SSDS signal around autosomal mo-2 regions was mostly independent of melanogaster SCs suggests that sister chromatids have structurally PRDM9 (Fig. 5G and fig. S8B). We conclude that autosomal mo-2 distinct axes during pachynema (47). Moreover, splitting of sister regions not only accumulate PAR-like levels of RMMAI proteins and chromatid axes occurs in cohesion-defective mutants of mouse (Rec8–/–) undergo PAR-like axis remodeling in spermatocytes, they frequently and Sordaria macrospora (spo76) (48, 49). There is thus ample precedent form DSBs in a PAR-like manner. for individualization of sister chromatids into separate axes, but to our knowledge, splitting of the PAR in late zygonema represents the earliest Behavior of mo-2-containing regions in oocytes stage yet documented for this phenomenon in wild-type in any In females, recombination between the two X chromosomes is not species. restricted to the PAR, so oocytes do not require the same high level of Previous studies showed that expressing only one of the two major PAR DSB formation as spermatocytes (43). We therefore asked whether splicing isoforms of Spo11 (Spo11β) or tagging SPO11 with the DNA the PAR undergoes spermatocyte-like structural changes in oocyte binding domain of yeast Gal4 confers a specific defect in PAR DSB meiosis. We detected robust accumulation of RMMAI proteins on mo-2- formation (7, 50). Our findings raise the possibility that the specialized containing regions in the PARs and ends in oocytes from properties of the PAR uniquely sensitize it to otherwise subtle defects in leptonema to pachynema (Fig. 6A,B and fig. S9A), also consistent with SPO11 activity. conventional micrographs showing blobs of MEI4 and ANKRD31 (22, It remains to be determined how the unique loop–axis structure of 31). Similar to spermatocytes, X PARs in oocytes displayed an extended the PAR emerges from RMMAI accumulation and how this structure is axis and compact PARb FISH signal from leptonema to zygonema and a related to DSB formation. Interestingly, budding yeast also uses early, transition to a shorter axis and more extended PARb signal in pachynema, robust recruitment of Rec114 and Mer2 (the IHO1 ortholog) as a means with gradual loss of REC114 signal upon synapsis (Fig. 6A). However, to ensure that its smallest chromosomes incur DSBs in every meiosis we were unable to detect obvious thickening (conventional microscopy) (51). Thus, even if the cis-acting determinants differ between species, or splitting (SIM) of the PAR axis, and REC8 did not accumulate to high such preferential recruitment appears to be an evolutionarily recurrent levels (Fig. 6A,B). Moreover, like the PAR (43), autosomal mo-2 regions strategy for mitigating the risk of recombination failure and showed little enrichment for DSB signal in SSDS data from wild-type missegregation that arises when the length of chromosomal homology is ovaries (fig. S8A and S9B). These findings suggest that hyper- limited. accumulation of (at least some of) the RMMAI proteins is an intrinsic feature of mo-2-associatd regions but also show that the full suite of Materials and Methods PAR(-like) structural changes and DSB formation is dependent on cellular context. Oocytes presumably lack one or more critical protein Mice factors or post-translational modifications. Mice were maintained and sacrificed under U.S.A. regulatory standards and experiments were approved by the Memorial Sloan Discussion Kettering Cancer Center (MSKCC) Institutional Animal Care and Use Committee. Animals were fed regular rodent chow with ad libitum access We demonstrate that the PAR in male mice undergoes a striking to food and water. The Ankrd31 knockout allele (Ankrd31em1Sky) is a rearrangement of loop–axis structure prior to DSB formation and single base insertion (+A) in exon 3; its generation and homologous pairing involving recruitment of RMMAI proteins, dynamic phenotypic characterization is described elsewhere (30). Mice with the axis elongation, and splitting of sister chromatid axes (fig. S10). Some Mei4 knockout allele (22) were kindly provided by B. de Massy (IGH, but not all of these behaviors also occur in oocytes. It appears that the Montpellier, France). All other mice were purchased from the Jackson mo-2 minisatellite array is a key cis-acting determinant and at least some Laboratory: C57BL/6J (stock #00664), MSM/MsJ (stock #003719), of the RMMAI proteins are crucial trans-acting determinants, with B6N(Cg)-Syce1tm1b(KOMP)Wtsi/2J (stock #026719), B6;129S7- ANKRD31 occupying an especially mo-2-specific (as opposed to global) Hormad1tm1Rajk/Mmjax (stock #41469-JAX), B6;129S4- role. Our findings provide mechanistic and ultrastructural explanation for Rec8mei8/JcsMmjax (stock #34762-JAX). Mice were genotyped using the long axes and short loops in the PAR (7). These PAR behaviors Direct Tail lysis buffer (Viagen) following the manufacturer’s appear to be essential for meiotic pairing, recombination, and segregation instructions. of heteromorphic sex chromosomes. RMMAI enrichment could occur through direct DNA binding by Generation of REC8 and REC114 antibodies one or more RMMAI proteins to the mo-2 DNA sequence (and/or another To produce antibodies against REC8, a fragment of the mouse Rec8 tightly linked DNA element) or to an mo-2-associated chromatin encoding amino acids 36 to 253 (NCBI Reference Sequence: structure. MEI4, REC114, and ANKRD31 (but not IHO1) are critical for NP_001347318.1) was cloned into pGEX-4T-2 vector. The resulting formation of RMMAI blobs and downstream structural changes [this fusion of the REC8 fragment fused to glutathione S tranferase (GST) was study and (31)], but ANKRD31 is less critical than the others for the expressed in E. coli, affinity purified on glutathione Sepharose 4B, and smaller, more widely distributed RMMAI foci that are thought to be cleaved with Precision protease. Antibodies were raised in rabbits by responsible for generating most DSBs genome wide (21-24). Ankyrin Covance Inc. (Princeton NJ) against the purified recombinant REC8 repeat domains—from which ANKRD31 derives its name—are often fragment, and antibodies were affinity purified using GST-REC836-253 involved in protein-protein interactions (44), including with chromatin that had been immobilized on glutathione sepharose by crosslinking with (45). The hyperaccumulation of RMMAI may thus reflect the fact that dimethyl pimelimidate; bound antibodies were eluted with 0.1 M glycine, mo-2 repeats provide a highly multivalent moiety that may be recognized pH 2.5. Purified antibodies were tested in western blots of testis extracts principally by ANKRD31 bound to REC114–MEI4 complexes. and specificity was validated by immunostaining of spread meiotic However, reliance on this repetitive structure likely imposes substantial chromosomes from wild type and Rec8–/– mice. instability through unequal exchange (36, 46). The PAR DNA structure To produce antibodies against REC114, a fragment of the mouse thus paradoxically is necessary for genome stability by supporting sex Rec114 gene encoding a truncated polypeptide lacking the N-terminal chromosome segregation but also promotes the known rapid evolution of 110 amino acids (NCBI Reference Sequence: NP_082874.1) was cloned mammalian PARs (6).

4 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Acquaviva et al., 30 Jan, 2019 – bioRxiv preprint v1 into pET-19b expression vector. The resulting hexahistidine-tagged or for at least 3 hours at room temperature. Slides were washed 3 ´ 5 min REC114111-259 fragment was insoluble when expressed in E. coli, so the in 1´ PBS, 0.05% Tween-20, then blocked for 10 min, and incubated with recombinant protein was solubilized and affinity purified on Ni-NTA secondary antibody for 1–2 hours at 37°C in a humid chamber. Slides resin in the presence of 8 M urea. Eluted protein was dialyzed against 100 were washed 3 ´ 5 min in the dark on a shaker with 1´ PBS, 0.05% mM NaH2PO4, 10 mM Tris-HCl, 6 M urea, pH 7.3 and used to immunize Tween-20, rinsed in H2O, and mounted before air drying with rabbits (Covance Inc.). Antibodies were affinity purified against purified Vectashield (Vector Labs). Antibody dilutions were centrifuged at recombinant His6-REC114111-259 protein immobilized on cyanogen 13,000 rpm for at least 5 min before use. Primary antibodies used were bromide-activated sepharose and eluted in 0.2 M glycine pH 2.5. The rabbit and guinea pig anti-ANKRD31 (30) (1:200 dilution), rabbit anti- affinity purified antibodies were previously used by Stanzione et al. (23), HORMAD2 (Santa Cruz, sc-82192, 1:50), guinea pig anti-HORMAD2 who reported detection of a band of appropriate molecular weight in (1:200) and guinea pig anti-IHO1 (1:200) (gifts from A. Toth (Technical western blots of testis extracts. However, subsequent analysis showed University of Dresden)), goat anti-MEI1 (Santa Cruz, sc-86732, 1:50), that this band is also present in extracts of Rec114–/– testes, and thus is rabbit anti-MEI4 (gift from B. de Massy, 1:200), rabbit anti-REC8 (this non-specific (C. Brun and B. de Massy, personal communication). study, 1:100), rabbit anti-REC114 (this study, 1:200), rabbit anti-RPA2 Importantly, however, Stanzione et al. also reported detection of (Santa Cruz, sc-28709, 1:50), goat anti-SYCP1 (Santa Cruz, sc-20837, immunostaining foci on spread meiotic chromosomes similar to findings 1:50), rabbit anti-SYCP2 (Atlas Antibodies, HPA062401, 1:100), mouse reported here and by Boekhout et al. (30). This immunostaining signal is anti-SYCP3 (Santa Cruz, sc-74569, 1:100), goat anti-SYCP3 (Santa absent from chromosome spreads prepared from Rec114–/– mutant mice Cruz, sc-20845, 1:50), rabbit anti-TRF1 (Alpha Diagnostic, TRF12-S, (C. Brun and B. de Massy, personal communication). Moreover, this 1:100). Secondary antibodies used were CF405S anti-guinea pig immunostaining signal is indistinguishable from that reported using (Biotium, 20356), CF405S anti-rabbit (Biotium, 20420), CF405S anti- independently generated and validated anti-REC114 antibodies (24). We mouse (Biotium, 20080), 488 donkey anti-mouse (Life technologies, conclude that our anti-REC114 antibodies are highly specific for the A21202), 488 donkey anti-rabbit (Life technologies, A21206), 488 cognate antigen when used for immunostaining of meiotic chromosome donkey anti-goat (Life technologies, A11055), 488 donkey anti-guinea spreads. pig (Life technologies, A11073), 568 donkey anti-mouse (Life technologies, A10037), 568 donkey anti-rabbit (Life technologies, Chromosome spreads A10042), 568 goat anti-guinea pig (Life technologies, A11075), 594 Testes were dissected and deposited after removal of the tunica donkey anti-mouse (Life technologies, A21203), 594 donkey anti-rabbit albuginea in 1´ PBS pH 7.4. Seminiferous tubules were minced using (Life technologies, A21207), 594 donkey anti-goat (Life technologies, forceps to form a cell suspension. The cell suspension was filtered A11058), 647 donkey anti-rabbit (Abcam, ab150067), 647 donkey anti- through a 70-µm cell strainer into a 15 ml Falcon tube pre-coated with goat (Abcam, ab150131), all at 1:250 dilution. 3% (w/v) BSA, and was centrifuged at 1000 rpm for 5 min. The cell pellet was resuspended in 12 ml of 1´ PBS for an additional centrifugation step ImmunoFISH and DNA probe preparation at 1000 rpm for 5 min and the pellet was resuspended in 1 ml of hypotonic All steps were performed in the dark to prevent loss of fluorescence buffer containing 17 mM sodium citrate, 50 mM sucrose, 30 mM Tris- from prior immunostaining. After the last washing step in the HCl pH 8, 5 mM EDTA pH 8, 0.5 mM dithiothreitol (DTT), 10 µl of immunostaining protocol, slides were placed horizontally in a humid 100´ Halt protease inhibitor cocktail (Thermo Scientific), and incubated chamber and the chromosome spreads were re-fixed with an excess of for 8 min. Next, 9 ml of 1´ PBS was added and the cell suspension was 2% (w/v) paraformaldehyde in 1´ PBS (pH 9.3) for 10 min at room centrifuged at 1000 rpm for 5 min. The cell pellet was resuspended in 100 temperature. Slides were rinsed once in H2O, washed for 4 min in 1´ mM sucrose pH 8 to obtain a slightly turbid cell suspension, and PBS, sequentially dehydrated with 70% (v/v) ethanol for 4 min, 90% incubated for 10 min. Superfrost glass slides were divided into two ethanol for 4 min, 100% ethanol for 5 min, and air dried vertically for 5- squares using an ImmEdge hydrophobic pen (Vector Labs), then 110 µl 10 min. Next, 15 µl of hybridization mix was applied containing the DNA of 1% paraformaldehyde (PFA) (freshly dissolved in presence of NaOH probe(s) in 70% (v/v) deionized formamide (Amresco), 10% (w/v) at 65°C, 0.15% Triton, pH 9.3, cleared through 0.22 µm filter) and 30 µl dextran sulfate, 2´ SSC buffer (saline sodium citrate), 1´ Denhardt’s of cell suspension was added per square, swirled three times for buffer, 10 mM EDTA pH 8 and 10 mM Tris-HCl pH 7.4. Cover glasses homogenization, and the slides were placed horizontally in a closed (22 x 22 mm) were applied and sealed with rubber cement (Weldwood humid chamber for 2 h. The humid chamber was opened for 1 h to allow contact cement), then the slides were denatured on a heat block for 7 min almost complete drying of the cell suspension. Slides were washed in a at 80°C, followed by overnight incubation (>14 h) at 37°C. Cover glasses Coplin jar 2 ´ 5 min in 1´ PBS on a shaker, and 2 min with 0.4% Photo- were carefully removed using a razor blade, slides were rinsed in 0.1´ Flo 200 solution (Kodak), air dried and stored in aluminum foil at –80°C. SSC buffer, washed in 0.4´ SSC, 0.3% NP-40 for 5 min, washed in PBS– Ovaries were extracted from 14.5–18.5 d post-coitum mice, and 0.05% Tween-20 for 3 min, rinsed in H2O, and mounted with Vectashield collected in 1´ PBS pH 7.4. After 15 min incubation in hypotonic buffer, before air drying. the ovaries were placed on a slide containing 30 µl of 100 mM sucrose To generate FISH probes, we used the nick translation kit from pH 8, and dissected with forceps to form a cell suspension. The remaining Abbott Molecular following the manufacturer’s instructions and using tissues were removed, 110 µl of 1% paraformaldehyde-0.15% Triton was CF dye-conjugated dUTP (Biotium), on BAC DNA from the clones added, and the slides were gently swirled for homogenization, before RP24-500I4 (maps to the region of the PAR boundary, PARb probe) incubation in a humid chamber as described above for spermatocyte CH25-592M6 (maps to the distal PAR, PARd probe), RP23-346H16, chromosome spreads. RP24-136G21, and CH36-200G6 (centromere-distal ends of chr4, chr9, and chr13, respectively). BAC clones were obtained from the BACPAC Immunostaining Resource Center (CHORI). Labeled DNA (500 ng) was precipitated Slides of meiotic chromosome spreads were blocked for 30 min at during 30 min incubation at -20°C after adding 5 µl of mouse Cot-1 DNA room temperature horizontally in a humid chamber with an excess of (Invitrogen), 0.5 volume of 7.5 M ammonium acetate and 2.5 volumes of blocking buffer containing 1´ PBS, pH 7.4 with 0.05% Tween-20, 7.5% cold 100% ethanol. After washing with 70% ethanol and air drying in the (v/v) donkey serum, 0.5 mM EDTA, pH 8.0, and 0.05% (w/v) sodium dark, the pellet was dissolved in 15 µl of hybridization buffer. azide, and cleared by centrifugation at 13,000 rpm for 15 min. Slides were Mo-2 oligonucleotide probes were synthetized by Integrated DNA incubated with primary antibody overnight in a humid chamber at 4°C, Technologies, with 6-FAM or TYE™ 665 fluorophores added to both 5¢

5 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Sex chromosome recombination and 3¢ ends of the oligonucleotide. The DNA sequence was designed chr13, chr9, and the PAR (two signals for each when unpaired, or a single based on the previously defined consensus sequence (37), and the probe signal for each after homologous pairing/synapsis). Notably, the RPA2 was used at a final concentration of 10 pmol/µl in hybridization buffer focus was most often found in a slightly more centromere-proximal without Cot-1 DNA. The Y-chromosome paint probe was purchased position compared to the bulk of mo-2 FISH signals, and therefore from IDLabs and used at 1:30 dilution in hybridization buffer without colocalized partly with mo-2 FISH signals. In the case of the PAR, this Cot-1 DNA. position corresponds closely to the region of the PAR boundary (PARb probe). Similar trend was observed on autosomal mo-2 clusters. EdU incorporation For estimates of chromatin extension, we measured the maximal Seminiferous tubules were incubated in DMEM with 10% FCS and axis-orthogonal distance between the FISH signal and the center of the 10 µM EdU at 37°C for 1 h for in vitro labeling. EdU incorporation was PAR axis, or the centromere-distal axis for chr9 stained by SYCP3. In detected using the Click-iT EdU Alexa Fluor 647 imaging kit mutant mice defective for RMMAI protein recruitment in the mo-2 (Invitrogen) according to the manufacturer's instructions. regions, the PAR axis was defined as the nearest SYCP3 segment adjacent to the telomeric SYCP3 signal. Image acquisition For quantification of RPA2, ANKRD31, REC8, and mo-2 signal Images of spread spermatocytes were acquired on a Zeiss Axio intensity in B6 ´ MSM and MSM ´ B6 F1 hybrids, late zygotene Observer Z1 Marianas Workstation, equipped with an ORCA-Flash 4.0 spermatocytes with at least one RPA2 focus on X or Y PAR were camera and DAPI, CFP, FITC, TEXAS red and Cy5 filter sets, analyzed. We used the elliptic selection tool in Fiji to define a region of illuminated by an X-Cite 120 PC-Q light source, with either 63´/1.4 NA interest around the largest signal in the PAR, and the same selection tool oil immersion objective or 100´/1.4 NA oil immersion objective. was then positioned on the other PAR axis for comparison. The Marianas Slidebook (Intelligent Imaging Innovations) software was used fluorescence intensity was measured as the integrated density with for acquisition. background substraction. Structured illumination microscopy (3D-SIM) was performed at the Bio-Imaging Resource Center in Rockefeller University using an OMX Prophase I sub-staging and identification of the PAR Blaze 3D-SIM super-resolution microscope (Applied Precision), Nuclei were staged according to the dynamic behavior of the equipped with 405 nm, 488 nm and 568 nm lasers, and 100´/1.40 NA autosome and sex chromosome axes during prophase I, using SYCP3 UPLSAPO oil objective (Olympus). Image stacks of several µm staining. Leptonema was defined as having short stretches of SYCP3 but thickness were taken with 0.125 µm z-steps, and were reconstructed no evidence of synapsis, early/mid-zygonema as having longer stretches in Deltavision softWoRx 6.1.1 software with a Wiener filter of 0.002 of SYCP3 staining and some synapsis, and late zygonema as having fully using wavelength specific experimentally determined OTF functions. assembled chromosome axes and substantial (>70%) synapsis. The X and Slides were prepared and stained as described above, except that Y chromosomes generally can be identified at this stage, and the PAR chromosomes were spread only on the central portion of the slides, and axis is distinguishable because it appears thicker than the centromeric the slides mounted using 18 ´ 18 mm coverslips (Zeiss). end, particularly near the end of zygonema when autosomes are almost fully synapsed. Early pachynema was defined as complete autosomal Image analysis synapsis, whereas the X and Y chromosomes could display various 3D-SIM images are shown either as a z-stack using the sum slices configuration: i) unsynapsed, with thickened PAR axes, ii) engaged in function in Fiji/ImageJ, or as a unique slice. The X and/or Y PAR synapsis, iii) synapsed in the PAR and non-homologously synapsed chromosomes were cropped, rotated and further cropped for best display. along the full (or nearly full) axis. Mid pachynema was For montage display, the X and Y chromosomes images were positioned defined as showing bright signal from autosome axes, desynapsing X and on a black background using Adobe Illustrator. In the instances where the Y axes remaining synapsed only in the PAR, with short PAR axis. During axes of the X and Y chromosomes were cropped, the area of cropping this stage, the autosomes and the non-PAR X and Y axes are initially was labeled with a light gray dotted line. Loop/axis measurements, foci short and thick, and progressively become longer and thinner. Late counts, and fluorescence intensity quantification were only performed on pachynema was defined as brighter autosome axes with a characteristic images from conventional microscopy using the original, unmodified thickening of all autosome ends. The X and Y non-PAR axes are then data. long and thin and show excrescence of axial elements. Diplonema was To measure the colocalization between RMMAI proteins, we defined as brighter axes and desynapsing autosome, associated with costained for SYCP3 and ANKRD31 along with either MEI4, REC114, prominent thickening of the autosome ends, particularly the centromeric or IHO1, and manually counted the number of ANKRD31 foci ends. In early diplonema, the non-PAR axes of X and Y chromosomes overlapping with SYCP3 and colocalizing or not with MEI4, REC114 or are still long and thin and progressively condense to form bright axes, IHO1. These counts were performed in 16 spermatocytes from leptonema associated with bulges. Most experiments were conducted using SYCP3 to early/mid zygonema. in combination with a RMMAI protein, which allows easier distinction To quantify the total number of RPA2, MEI4, REC114, ANKRD31, between synapsing and desynapsing X and Y chromosomes. and IHO1 foci, single cells were manually cropped and analyzed with By using only SYCP3 staining, the PARs can only be identified semi-automated scripts in Fiji (52) as described in detail elsewhere (30). unambiguously from the late zygonema-to-early pachynema transition Briefly, images were auto-thresholded on SYCP3 staining, which was through to diplonema. From pre-leptonema to mid/late-zygonema, the used as a mask to use ‘Find Maxima’ to determine the number of foci. PARs were identified as the two brightest RMMAI signals, the two Images were manually inspected to determine that there were no obvious brightest mo-2 FISH signals, the two brightest PARb FISH signals, or the defects in determining SYCP3 axes, that no axes from neighboring cells two FISH signals from the PARd probe. The Y PAR could be were counted, that artifacts were present, or that foci were missed by the distinguished from the X PAR using the PARb probe, as this probe also script. weakly stains the chromatin of the non-PAR portion of the Y To test for colocalization between RPA2 and mo-2 FISH signals, we chromosome. manually scored the percentage of mo-2 FISH signals colocalizing at PAR loop/axis measurements in oocytes were performed on two least partly with RPA2. Depending on the progression of synapsis during 14.5–15.5 dpc (days post-coitum) (enriched for leptotene and zygotene prophase I, between eight and four discrete mo-2 FISH signals could be oocytes) and two 18.5 dpc female fetuses (enriched for pachytene detected, corresponding to (with increasing signal intensity) the chr4, oocytes).

6 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Acquaviva et al., 30 Jan, 2019 – bioRxiv preprint v1

We found significant variability in the X or Y PAR axis length References between different animals in our mouse colony maintained in a C57BL/6J 1. D. Zickler, N. Kleckner, Meiotic chromosomes: integrating structure and congenic background, and even between different C57BL/6J males function. Annu Rev Genet 33, 603-754 (1999). obtained directly from the Jackson Laboratory. This is in agreement with 2. S. Keeney, J. Lange, N. Mohibullah, Self-organization of meiotic previous reports about the hypervariable nature of the mo-2 minisatellite recombination initiation: general principles and molecular pathways. Annu and its involvement in unequal crossing over in the mouse (4, 37, 46, 53, Rev Genet 48, 187-214 (2014). 54) (mo-2 was also named DXYmov15 or Mov15 flanking sequences). 3. S. Palmer, J. Perry, D. Kipling, A. Ashworth, A gene spans the However, the RMMAI signal intensity/elongation and the PAR axis pseudoautosomal boundary in mice. Proc Natl Acad Sci U S A 94, 12030- length were always correlated with mo-2 FISH signal intensity. 12035 (1997). Importantly, despite this variability, mo-2 and RMMAI proteins were 4. P. Soriano et al., High rate of recombination and double crossovers in the enriched in the PAR and autosome ends of all mice analyzed. mouse pseudoautosomal region during male meiosis. Proc Natl Acad Sci U S A 84, 7218-7220 (1987). Analysis of SSDS data 5. J. Perry, S. Palmer, A. Gabriel, A. Ashworth, A short pseudoautosomal SSDS sequencing data were from previously described studies (9, region in laboratory mice. Genome Res 11, 1826-1832 (2001). 30, 43) and are all available at the Gene Expression Omnibus (GEO) 6. T. Raudsepp, P. J. Das, F. Avila, B. P. Chowdhary, The pseudoautosomal repository under accession numbers GSE35498, GSE99921, region and sex chromosome in domestic species. Sex Dev 6, 72- GSE118913. To define enrichment values presented in fig. S8A, the 83 (2012). SSDS coverage was summed across the indicated coordinates adjacent to 7. L. Kauppi et al., Distinct properties of the XY pseudoautosomal region the mo-2 repeats. A chromosomal mean and standard deviation for chr9 crucial for male meiosis. Science 331, 916-920 (2011). was estimated by dividing the chromosome into 4-kb bins, summing the 8. L. Kauppi, M. Jasin, S. Keeney, The tricky path to recombining X and Y SSDS coverage in each bin, and calculating the mean and standard chromosomes in meiosis. Ann N Y Acad Sci 1267, 18-23 (2012). deviation after excluding those bins that overlapped a DSB hotspot. The enrichment score was then defined as the difference between the 9. K. Brick, F. Smagulova, P. Khil, R. D. Camerini-Otero, G. V. Petukhova, is directed away from functional genomic elements coverage in the mo-2-adjacent region and the chr9 mean coverage, in mice. Nature 485, 642-645 (2012). divided by the chr9 standard deviation. 10. J. Lange et al., The landscape of mouse meiotic double-strand break Statistical analysis formation, processing, and repair. Cell 167, 695-708 e616 (2016). All statistical tests were performed in R (version 3.4.4) (55) and 11. N. Kleckner, Chiasma formation: chromatin/axis interplay and the role(s) of RStudio (Version 1.1.442). Negative binomial regression was calculated the synaptonemal complex. Chromosoma 115, 175-194 (2006). using the glm.nb function from the MASS package (version 7.3-49) (56). 12. Y. Blat, R. U. Protacio, N. Hunter, N. Kleckner, Physical and functional interactions among basic chromosome organizational features govern early Data and software availability steps of meiotic chiasma formation. Cell 111, 791-802 (2002). Image analysis scripts are available on Github: 13. S. Panizza et al., Spo11-accessory proteins link double-strand break sites to https://github.com/Boekhout/ImageJScripts. the chromosome axis in early meiotic recombination. Cell 146, 372-383 (2011). End Matter 14. C. Heyting et al., Identification of two major components of the lateral elements of synaptonemal complexes of the rat. Eur J Cell Biol 43, 148-154 (1987). Author Contributions and Notes LA designed and conducted all of the cytogenetic experiments presented 15. J. A. Schalk et al., Localization of SCP2 and SCP3 protein molecules within and analyzed the data. MEK generated Ankrd31 mutant mice and anti- synaptonemal complexes of the rat. Chromosoma 107, 540-548 (1998). ANKRD31 antibodies. MB and MEK provided Ankrd31 mutant mice and 16. S. Rankin, Complex elaboration: making sense of meiotic cohesin dynamics. unpublished data. KB and FP performed SSDS and analyzed the data FEBS J 282, 2426-2443 (2015). under the supervision of RDC with input from LA and SK. MvO 17. T. Fukuda, K. Daniel, L. Wojtasz, A. Toth, C. Hoog, A novel mammalian generated REC8 and REC114 antibodies. LK performed initial HORMA domain-containing protein, HORMAD1, preferentially associates characterization and provided unpublished data on PAR ultrastructure with unsynapsed meiotic chromosomes. Exp Cell Res 316, 158-171 (2010). and cohesin enrichment. MJ and SK designed and supervised the 18. L. Wojtasz et al., Mouse HORMAD1 and HORMAD2, two conserved research, analyzed data, and secured funding. LA and SK wrote the meiotic chromosomal proteins, are depleted from synapsed chromosome manuscript with input from MJ. All authors edited the manuscript. The axes with the help of TRIP13 AAA-ATPase. PLoS Genet 5, e1000702 authors declare no conflict of interest. (2009). 19. K. Daniel et al., Meiotic homologue alignment and its quality surveillance Acknowledgments are controlled by mouse HORMAD1. Nat Cell Biol 13, 599-610 (2011). We thank Attila Tóth (Technische Universität Dresden, Germany) and 20. Y. H. Shin et al., Hormad1 mutation disrupts synaptonemal complex Bernard de Massy (Institut Génétique Humaine, Montpellier, France) for formation, recombination, and chromosome segregation in mammalian antibodies, mice, discussions, and sharing of unpublished information. meiosis. PLoS Genet 6, e1001190 (2010). We thank Alison North and the Bio-Imaging Resource Center at 21. R. Kumar et al., MEI4 - a central player in the regulation of meiotic DNA Rockefeller University for assistance with SIM. This work utilized the double-strand break formation in the mouse. J Cell Sci 128, 1800-1811 computational resources of the NIH HPC Biowulf cluster (2015). (http://hpc.nih.gov). Funding: MSKCC core facilities are supported by 22. R. Kumar, H. M. Bourbon, B. de Massy, Functional conservation of Mei4 Cancer Center Support Grant P30 CA008748. LA was supported in part for meiotic DNA double-strand break formation from yeasts to mice. by a fellowship from the Lalor Foundation. MB was supported in part by Dev 24, 1266-1280 (2010). a Rubicon fellowship from the Netherlands Organization for Scientific 23. M. Stanzione et al., Meiotic DNA break formation requires the unsynapsed Research. MvO was supported in part by National Institutes of Health chromosome axis-binding protein IHO1 (CCDC36) in mice. Nat Cell Biol (NIH) fellowship F32 GM096692. This work was supported by NIGMS 18, 1208-1220 (2016). grants R35 GM118092 (SK) and R35 GM118175 (MJ). 24. R. Kumar et al., Mouse REC114 is essential for meiotic DNA double-strand

7 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Sex chromosome recombination

break formation and forms a complex with MEI4. Life Sci Alliance 1, of meiotic double-strand break initiation in mammals. Methods Enzymol e201800259 (2018). 601, 391-418 (2018). 25. I. Lam, S. Keeney, Mechanism and regulation of meiotic recombination 42. P. P. Khil, F. Smagulova, K. M. Brick, R. D. Camerini-Otero, G. V. initiation. Cold Spring Harb Perspect Biol 7, a016634 (2014). Petukhova, Sensitive mapping of recombination hotspots using sequencing- 26. B. de Massy, Initiation of meiotic recombination: how and where? based detection of ssDNA. Genome Res, (2012). Conservation and specificities among eukaryotes. Annu Rev Genet 47, 563- 43. K. Brick et al., Extensive sex differences at the initiation of genetic 599 (2013). recombination. Nature 561, 338-342 (2018). 27. S. L. Page, R. S. Hawley, The genetics and molecular biology of the 44. J. Li, A. Mahajan, M. D. Tsai, Ankyrin repeat: A unique motif mediating synaptonemal complex. Annu Rev Cell Dev Biol 20, 525-558 (2004). protein-protein interactions. Biochemistry 45, 15168-15178 (2006). 28. J. Page et al., Inactivation or non-reactivation: what accounts better for the 45. G. Saredi et al., H4K20me0 marks post-replicative chromatin and recruits silence of sex chromosomes during mammalian male meiosis? Chromosoma the TONSL-MMS22L DNA repair complex. Nature 534, 714-718 (2016). 121, 307-326 (2012). 46. K. Harbers, P. Soriano, U. Muller, R. Jaenisch, High frequency of unequal 29. L. G. Reinholdt, J. C. Schimenti, Mei1 is epistatic to Dmc1 during mouse recombination in pseudoautosomal region shown by proviral insertion in meiosis. Chromosoma 114, 127-134 (2005). transgenic mouse. Nature 324, 682-685 (1986). 30. M. Boekhout et al., REC114 partner ANKRD31 controls number, timing and 47. C. K. Cahoon et al., Superresolution expansion microscopy reveals the three- location of meiotic DNA breaks. bioRxiv DOI: dimensional organization of the Drosophila synaptonemal complex. Proc https://doi.org/10.1101/425322, (2018). Natl Acad Sci U S A 114, E6857-E6866 (2017). 31. F. Papanikos et al., ANKRD31 regulates spatiotemporal patterning of 48. A. Agostinho et al., High density of REC8 constrains sister chromatid axes meiotic recombination initiation and ensures recombination between and prevents illegitimate synaptonemal complex formation. EMBO Rep 17, heterologous sex chromosomes in mice. bioRxiv DOI: 901-913 (2016). https://doi.org/10.1101/423293, (2018). 49. A. Storlazzi et al., Meiotic double-strand breaks at the interface of 32. L. L. Tres, Side-by-side pairing of the XY bivalent in spermatocytes and the chromosome movement, chromosome remodeling and reductional division. ubiquity of the H-Y . Arch Androl 2, 101-107 (1979). Genes and Development 17, 2675-2687 (2003). 33. A. J. Solari, The spatial relationship of the X and Y chromosomes during 50. F. Smagulova et al., Suppression of genetic recombination in the meiotic prophase in mouse spermatocytes. Chromosoma 29, 217-236 (1970). pseudoautosomal region and at subtelomeres in mice with a hypomorphic 34. L. L. Tres, Extensive pairing of the XY bivalent in mouse spermatocytes as Spo11 allele. BMC Genomics 14, 493 (2013). visualized by whole-mount electron microscopy. J Cell Sci 25, 1-15 (1977). 51. H. Murakami, I. Lam, J. Song, M. van Overbeek, S. Keeney, Multilayered 35. L. Chong et al., A human telomeric protein. Science 270, 1663-1667 (1995). mechanisms ensure that short chromosomes recombine in meiosis. bioRxiv doi: https://doi.org/10.1101/406892, (2018). 36. K. Harbers, U. Francke, P. Soriano, R. Jaenisch, U. Muller, Structure and chromosomal mapping of a highly polymorphic repetitive DNA sequence 52. J. Schindelin et al., Fiji: an open-source platform for biological-image from the pseudoautosomal region of the mouse sex chromosomes. Cytogenet analysis. Nat Methods 9, 676-682 (2012). Cell Genet 53, 129-133 (1990). 53. D. Kipling et al., Structural variation of the pseudoautosomal region between 37. Y. Takahashi et al., Methylation imprinting was observed of mouse mo-2 and within inbred mouse strains. Proc Natl Acad Sci U S A 93, 171-175 macrosatellite on the pseudoautosomal region but not on . (1996). Chromosoma 103, 450-458 (1994). 54. D. Kipling, E. C. Salido, L. J. Shapiro, H. J. Cooke, High frequency de novo 38. L. A. Bannister, L. G. Reinholdt, R. J. Munroe, J. C. Schimenti, Positional alterations in the long-range genomic structure of the mouse cloning and characterization of mouse mei8, a disrupted allelle of the meiotic pseudoautosomal region. Nat Genet 13, 78-80 (1996). cohesin Rec8. Genesis 40, 184-194 (2004). 55. R Core Team, R: A language and environment for statistical computing., (R 39. Y. Costa et al., Two novel proteins recruited by synaptonemal complex Foundation for Statistical Computing, Vienna, Austria, 2018). protein 1 (SYCP1) are at the centre of meiosis. J Cell Sci 118, 2755-2762 56. W. N. Venables, B. D. Ripley, Modern Applied Statistics with S. (Springer, (2005). New York, ed. 4, 2002). 40. K. P. Kim et al., Sister cohesion and structural axis components mediate 57. A. J. Solari, Sex Chromosomes and Sex Determination in Vertebrates. (CRC homolog bias of meiotic recombination. Cell 143, 924-937 (2010). Press, Boca Raton, 1993). 41. K. Brick, F. Pratto, C. Y. Sun, R. D. Camerini-Otero, G. Petukhova, Analysis

8 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. 1: Ultrastructure of the PAR during male meiosis. (A) Representative images showing axis thickening (SYCP2 and SYCP3) and ANKRD31 accumulation on X and Y PARs (arrowheads) in late zygonema. Scale bar: 2 μm. (B) Colocalization of ANKRD31 and MEI4 in a representative zygotene spermatocyte. Arrowheads indicate blobs. Dashed boxes are shown at higher magnification at the right. Graph: total number of MEI4 and ANKRD31 foci colocalized in leptotene/zygotene spermatocytes. Underlying data for this an all other graphs are provided in Data File S1. Scale bar: 2 μm. (C) PARb FISH probe colocalizes with REC114 blobs. Scale bar: 2 μm. (D) Ultrastructure of the PAR before and after synapsis. A montage of representative SIM images is shown. Dashed lines indicate where chromosomes are cropped. Scale bar: 1 μm. (E) RMMAI proteins are enriched along the extended PAR axes. Scale bars: 1 µm. (F) Schematic of PAR ultrastructure and distribution of axis and RMMAI proteins at late zygonema. (G) Telomere-binding protein TRF1 decorates the tip of the PAR bubble, ruling out a foldback (crozier) configuration. Scale bar: 1 µm. (H) PAR axis differentiation occurs in the absence of SPO11-generated DSBs. Scale bar: 1 µm.

9 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. 2. Dynamic remodeling of the PAR loop–axis structure. (A) Time course of REC8 and ANKRD31 immunostaining along the PAR axis from pre-leptonema (preL, left) to mid pachynema (right). A montage of representative sex chromosome SIM images is shown. Chromosomes a–e are presumptive X or Y, but could instead be the distal end of chr9. Chromosomes at later stages were unambiguously identified by morphology. Chromosomes i–k show examples where the initial pairing (probably synaptic) contact between X and Y is (i) centromere-proximal (that is, closer to the PAR boundary), (k) distal (that is, closer to the telomere), or (j) interstitial. Scale bar: 1 μm. (B) Time course of the spatial organization of the PAR loop–axis ensemble. We collected three measurements from the indicated number (N) of conventional immuno-FISH images from leptonema through mid-pachynema: length of the REC114 signal along the PAR axis; maximal distance from the PARb FISH signal to the distal end of the SYCP3-defined axis; and axis-orthogonal extension of FISH signal for the PARb probe (a proxy for loop sizes). Insets show examples of each type of measurement at each stage. Horizontal black lines on graph indicate means, which are also given numerically below the graph. Quantification from additional mice is provided in fig. S3. Scale bar: 1 μm.

10 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Fig. 3. Arrays of the mo-2 minisatellite are sites of RMMAI protein enrichment in the PAR and on autosomes. (A) Top panel: Self alignment of the PARb FISH probe, showing its highly repetitive nature. The circled block in the dot plot is a 20-kb cluster of the 31-bp mo-2 minisatellite (37). Bottom panel: Schematic depicting the last 1.4 Mb of the non-centromeric ends of the indicated chromosomes, showing the presence of mo-2 repeats (green) adjacent to gaps in chromosome assemblies (mm10). Mo-2 repeats also appear at the distal end of the chr4 in the shotgun assembly from Celera (Mm_Celera, 2009/03/04). The positions of the BAC clones used for FISH are indicated (PARb and PARd). (B) Colocalization of REC114 blobs with FISH signal using the mo-2 minisatellite sequence as a probe. A representative zygotene spermatocyte is shown. Scale bar: 2 μm. (C) PAR enrichment for ANKRD31 and RPA2 correlates with mo-2 copy number. Top panels: representative micrographs of late zygotene spermatocytes from reciprocal F1 hybrid males from crosses of B6 (high mo-2 copy number) and MSM (low mo-2 copy number) parents. Scale bars: 1 µm. Bottom panels: quantification of PAR-associated signals (A.U., arbitrary units) on B6-derived chromosomes (XB and YB) and MSM-derived chromosomes (XM and YM) from the indicated number of spermatocytes (N). Red lines indicate means. Differences between X and Y PAR intensities are significant for both proteins and for mo-2 FISH in both F1 hybrids (p < 0.0001, paired t- test).

11 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. 4. Requirements for RMMAI recruitment and PAR axis remodeling. (A) Quantification of REC114, ANKRD31, MEI4, and IHO1 foci along unsynapsed axes in the indicated number of leptotene/early zygotene spermatocytes (N) in wild type and the indicated mutants. Horizontal lines indicate means. Statistical significance for each comparison to wild type is indicated (Student’s t test): * = p<0.02, ** = p≤10-7, ns = not significant (p>0.05). Representative micrographs of REC114 staining are shown; images for other proteins are in fig. S5A. The presence or absence of mo-2 associated blobs (orange arrowheads) is indicated in the bottom panel for each mutant. Scale bars: 2 μm. (B) Genetic requirements for PAR loop–axis organization. Using conventional immuno-FISH microscopy for REC114 and mo-2 in late-zygotene or late-zygotene-like spermatocytes, we measured the length of the REC114 signal along the PAR axis, the length of the mo-2 FISH signal along the PAR axis, and the extension of the mo-2 FISH signal orthogonal to the axis. Quantification from additional mice is provided in fig. S5B. (C) Representative SIM images showing the Y-PAR loop–axis structure in each mutant at late zygonema. Scale bar: 1 µm.

12 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. 5. PAR-like structural reorganization and DSB formation on autosomal mo-2 arrays. (A) The mo-2-containing portion (arrowheads) of chr9 undergoes axis elongation and splitting similar to PARs. A representative SIM image of a wild- type zygotene spermatocyte is shown. See fig. S6A for images of the entire spread and of the Y and PAR-containing part of the X for comparison. Scale bar: 1 µm. (B) Loop-axis organization of the mo-2 region of chr9 in wild type and Ankrd31–/– late- zygotene spermatocytes. Compare to related parameters for the PAR in Fig. 4B. Scale bars: 1 µm. (C–E) ANKRD31 is required for high-level DSB formation in mo-2 regions in the PAR and on autosomes. Immuno-FISH for RPA2 and mo-2 was used to detect DSBs cytologically in wild type and the indicated mutants. Panel C and fig. S7A show representative micrographs (scale bar: 2 µm, inset 1 µm). Panel D shows global counts of RPA2 foci for zygotene (zyg) or zygotene-like cells and for pachytene (pach) or pachytene-like cells. Red lines are means. Panel E quantifies mo-2-associated RPA2 foci in the

13 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. same cells scored in panel D: for each cell, the fraction of mo-2 regions that had a colocalized RPA2 focus was scored. Red lines are means. Statistical significance is indicated for comparisons (Student’s t tests) of wild type to Ankrd31–/– or of Syce1–/– to Rec8–/– or Hormad1–/– for matched stages. Note that the number of discretely scorable mo-2 regions in panel E varied from cell to cell depending on pairing status. (F) X–Y pairing status, quantified by immuno-FISH for SYCP3 and the PARb probe. (G) PAR-like DSB formation near autosomal mo-2 regions. SSDS sequence coverage (data from (9, 30)) is shown for the mo- 2-adjacent region of chr9. The dashed portion f indicates a gap in the sequence assembly. Positions of mo-2 repeats are shown below. Chr13 is in fig. S8B and the PAR is analyzed separately (30).

Fig. 6. PAR behavior in oocytes. (A) PAR ultrastructure in oocytes, quantified as in Fig. 2B. Insets show examples of each type of measurement at each stage; late zygotene cells with PAR synapsis are compiled separately from other zygotene cells. Horizontal black lines on graphs indicate means. Scale bar: 1 μm. (B) Representative SIM image of a wild-type late zygotene oocyte showing neither detectable splitting of the PAR axis nor REC8 enrichment. Scale bar 2 µm.

14 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Supplementary Materials for

Ensuring meiotic DNA break formation in the mouse pseudoautosomal region

Laurent Acquaviva, Michiel Boekhout, Mehmet E. Karasu, Kevin Brick, Florencia Pratto, Megan van Overbeek, Liisa Kauppi, R. Daniel Camerini-Otero, Maria Jasin and Scott Keeney

Correspondence to [email protected], [email protected] or [email protected]

This PDF file includes:

Figs. S1 to S10

Other Supplementary Materials for this manuscript include the following:

Data S1: Excel spreadsheet containing the underlying data for graphs in the figures.

1 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. S1. PAR axis thickening and accumulation of RMMAI proteins. (A) Axis thickening (SYCP3 and HORMAD2 staining) on the PAR (arrowhead) in a late zygotene spermatocyte. Scale bar: 2 μm. (B) Image adapted under Creative Commons CC-BY license from (18) showing enrichment of HORMAD1 on the thick PAR axis of the Y chromosome. (C) Colocalization of ANKRD31 and REC114, IHO1, and MEI1. Representative zygotene spermatocytes are shown. Arrowheads indicate densely staining blobs. Areas indicated by dashed boxes are shown at higher magnification at the right. The graphs show the total number of foci colocalized in leptotene/zygotene spermatocytes. N.D., not determined: The low immunofluorescence signal for MEI1 did not allow us to quantify the colocalization with ANKRD31, although MEI1 showed clear colocalization with ANKRD31 in the blobs and at least some autosomal foci (insets). Scale bars: 2 μm. Further evidence for extensive colocalization with ANKRD31 is documented in separate studies (30, 31). (D) ANKRD31, REC114, and MEI1 immunostaining starts to appear in pre-leptonema. Seminiferous tubules were cultured with 5-ethynyl-3¢-deoxyuridine (EdU) to label replicating cells, then chromosome spreads were stained for SYCP3 and either MEI1 plus REC114 or ANKRD31 plus PARb FISH. Colocalized foci appear in pre- leptonema (EdU-positive cells that are weakly SYCP3-positive), as previously shown for MEI4 and IHO1 (21, 23). Because we can already detect ANKRD31 accumulation at sites of PARb-hybridization, we infer that the stronger sites of accumulation of MEI1 and REC114 also include PARs. Scale bars: 2 μm. 2 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. S2. PAR ultrastructure. (A) Comparison of conventional microscopy and SIM, showing that the thickened PAR axis in conventional microscopy is resolved as separated axial cores (arrowheads). Scale bars: 2 μm. (B) Ultrastructure of axis proteins SYCP2, SYCP3, and HORMAD2 in the PAR. Scale bars: 1 µm. (C,D) Paired PARs with elongated and split axes occur in late zygonema to early pachynema. Shown are electron micrographs adapted with permission from (34) in comparison with SIM immunofluorescence images of spermatocytes at early pachynema (panel C; same chromosomes as in Fig. 1E) or late zygonema (panel D; cyan arrowheads indicate examples of incomplete autosomal synapsis). The spermatocytes in the electron micrographs were originally considered to be in mid-to-late pachynema (34). However, in our SIM experiments, we can only detect this structure (paired X and Y with elongated and split axes, resembling a crocodile’s jaws) around the zygotene–pachytene transition, when RMMAI proteins are still highly abundant on the PAR axes, and when most or all autosomes are completely synapsed. Moreover, other published electron micrographs from mid-to-late pachytene spermatocytes show diagnostic ultrastructural features that are not present in the electron micrographs reproduced here, including a short PAR axis length, multi- stranded stretches of axis on non-PAR portions of the X and Y chromosomes with excrescence of axial elements, and a clear thickening of autosomal (28, 57). These observations allow us to conclude definitively that the elongation and splitting of PAR axes are a hallmark of cells from late zygonema into early pachynema. Scale bars in SIM images: 1 μm in panel C, 2 µm in panel D. (E) Cartoon of hypothetical crozier configuration in which a single conjoined axis for both sister chromatids is folded back on itself. (F) Separation of PAR sister chromatid axes. FISH signal for the PARb probe is arrayed relatively symmetrically on both axial cores, consistent with separated sister chromatid axes (a bubble configuration). Scale bar: 1 µm.

3 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. S3. Time course of the spatial organization of the PAR loop–axis ensemble. Measurements of REC114 spreading along the PAR axis, length of the PAR axis, and extension of PARb chromatin orthogonal to the axis were collected as in Fig. 2B, on two additional males. Data from mouse 2 are reproduced from Fig. 2B to facilitate comparison. Means of each measurement for each mouse at each stage are given below, along with the means across all three mice. Means are rounded to two significant figures; the grand mean was calculated using unrounded values from individual mice. The number of cells of each stage from each mouse is given (N). Modest variability in the apparent dimensions of the Y chromosome PAR between different mice may be attributable to variation in copy number of mo-2 and other repeats because of unequal exchange during meiosis. Nonetheless, highly similar changes in spatial organization over time in prophase were observed in all mice examined, namely progressive elongation then shortening of axes and concomitant lengthening of loops.

4 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

5 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Fig. S4. RMMAI enrichment at mo-2 minisatellite arrays in the PAR and on specific autosomes. (A) Comparison of mo-2 FISH with REC114 localization relative to the PAR boundary (PARb FISH probe) and the distal PAR (PARd probe). In mid zygonema, the mo-2 FISH signal colocalizes well with REC114 staining in between the PARb and PARd FISH signals. In late zygonema, mo-2 and REC114 are similar to one another and are elongated along the thickened SYCP3 staining of the PAR axis. From early to mid pachynema, REC114 progressively disappears, whereas the mo-2 FISH signal becomes largely extended away from the PAR axes. Note that the relative positions of the PARb and PARd probes reinforce the conclusion that the PAR does not adopt a crozier configuration. Scale bar: 1 μm. (B) Illustration of the compact organization of the PAR chromatin (mo-2 FISH signal) compared to a whole-Y-chromosome paint probe. Scale bar: 2μm. (C) Confirmation that autosomal mo-2 FISH signals match the chromosomal locations indicated by mm10 or Celera genome assemblies. FISH was performed using an oligonucleotide probe containing the mo-2 consensus sequence in combination with BAC probes for adjacent segments of chromosomes 13, 9 and 4, as indicated. Magenta arrows point to concordant FISH signals. The chr13 BAC probe also hybridizes to the PAR. Scale bars: 2 µm. (D) Lower mo-2 copy number in the M. m. molossinus subspecies correlates with lower REC114 staining in mo-2 regions. The top panels compare MSM and B6 mice for the colocalization between REC114 immunostaining and mo-2 FISH in leptotene spermatocytes. The REC114 and SYCP3 channels are shown at equivalent exposure for the two strains, whereas a longer exposure is shown for the mo-2 FISH signal in the MSM spermatocyte. Note that the mo-2-associated REC114 blobs are much brighter relative to the smaller dispersed REC114 foci in the B6 spermatocyte than in MSM. The bottom panel shows representative pachytene spermatocytes to confirm the locations of mo-2 clusters at autosome ends and the PAR in the MSM background. Scale bars: 2 μm. (E) Representative micrographs of late zygotene spermatocytes from reciprocal F1 hybrid males from crosses of B6 (high mo-2 copy number) and MSM (low mo-2 copy number) parents. Scale bar: 1 µm. (F) Frequency of paired X and Y at late zygonema and mid pachynema analyzed in three MSM and three B6 males. Differences between strains were not statistically significant at either stage (p > 0.13, Student’s t test). Note also that MSM X and Y are late pairing chromosomes, as in the B6 background. The similar pairing kinetics indicates that the lower intensity of RMMAI staining on the MSM PAR is not attributable to earlier PAR pairing and synapsis in this strain. The number of spermatocytes analyzed is indicated (N).

6 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

7 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Fig. S5. Genetic requirements for RMMAI assembly on chromosomes and for PAR loop–axis organization. (A) Representative micrographs of ANKRD31, MEI4, IHO1 and MEI1 staining in wild type and the indicated mutants (quantification is in Fig. 4A). Scale bars: 2 μm. (B) Measurements of PAR loop–axis organization, as in Fig. 4B, on two additional males. Data from mouse 1 are reproduced from Fig. 4B to facilitate comparison. Means of each measurement for each mouse at each stage are given below, along with the means across all three mice. Means are rounded to two significant figures; the grand mean was calculated using unrounded values from individual mice. The number of cells of each stage from each mouse is given (N). (C) REC8 is dispensable for splitting apart of PAR sister chromatid axes, but is required to maintain the connection between sisters at the distal tip of the chromosome. A representative SIM image is shown of a Y chromosome from a late zygotene Rec8–/– mutant spermatocyte. The SYCP3-labeled axes adopt an open fork configuration. Note that the distal FISH probe (PARd) shows that there are clearly disjoined sisters whereas the PAR boundary (PARb) shows only a single compact signal comparable to wild type. The disposition of the probes and SYCP3 further rules out the crozier configuration as an explanation for split PAR axes. Scale bar: 1 µm. (D) Quantification of REC114 and MEI4 foci in two additional pairs of wild type and Ankrd31–/– mice. Horizontal lines indicate means. Fewer foci were observed in the Ankrd31–/– mutant (p < 0.02, Student’s t test) for each comparison of mutant to wild type. (E) Reduced REC114-staining intensity of axis-associated foci in Ankrd31–/– mutants. To rigorously control for slide-to-slide and within- slide variation in immunostaining, we mixed together wild-type and Ankrd31–/– testis cell suspensions before preparing chromosome spreads. A representative image is shown of a region from a single microscopic field containing two wild-type zygotene spermatocytes (left) and two Ankrd31–/– spermatocytes of equivalent stage (right). Note the diminished intensity of REC114 foci in the Ankrd31–/– spermatocytes. Scale bar: 2 µm.

8 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. S6. RMMAI recruitment and axis remodeling are functionally connected to mo-2 presence and copy number. (A) The mo-2- containing portion of chr9 undergoes axis elongation and splitting similar to PARs. The upper panel shows the SIM image for the complete wild-type zygotene spermatocyte from which the higher magnification detail of chr9 (identified based on chromosome length and brightness of REC114 staining) is shown in Fig. 5A. Higher magnification views of the Y and PAR-containing part of the X are shown for comparison. Scale bar: 1 µm. (B) Low mo-2 copy number on M. m. molossinus sex chromosomes is associated with a lesser degree of loop–axis reorganization. Representative SIM images are shown of late zygotene spermatocytes from F1 hybrid males from reciprocal crosses of B6 (high mo-2 copy number) and MSM (low mo-2 copy number) parents. Scale bars: 1 µm.

9 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. S7. PAR-associated RPA2 foci. (A) Immuno-FISH for RPA2 and mo-2 was used to detect DSBs cytologically in wild type and the indicated mutants. Representative images are shown for additional mutants, supplementing the images in Fig. 5C. Quantification is in Fig. 5D,E. Scale bars: 2 µm. (B) Frequent DSB formation at mo-2 regions in the PARs and on autosomes does not require HORMAD1. Micrograph at left shows two adjacent spermatocytes (boundary indicated by dashed line). Insets at right show higher magnification views of the numbered mo-2 regions, all of which are associated with RPA2 immunostaining of varying intensity. Scale bar: 2 µm. (C) Montage of SIM images from a B6 male showing that multiple, distinct RPA2 foci can be detected from late zygonema to mid pachynema, suggesting that multiple PAR DSBs can be formed during one meiosis (see also (7) for further discussion). Scale bar: 1 μm. (D) Percentage of spermatocytes at the zygotene-pachytene transition with no (0), 1, 2 or 3 distinguishable RPA2 foci on the unsynapsed Y chromosome PAR of MSM and B6 mice. The difference between the strains is statistically significant (negative binomial regression, p = 7.2 ´ 10-5). N indicates the number of spermatocytes analyzed. A representative picture is shown for each genotype, with one RPA2 focus on the MSM PAR and two apparent sites of RPA2 accumulation on the B6 PAR. Scale bar: 1 μm.

10 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. S8. DSB maps on the PAR and autosomal mo-2 regions. (A) Regions adjacent to the mo-2 region on chr9 show SSDS signal that is reproducibly elevated relative to chr9 average in wild-type testis samples but not in maps from Ankrd31–/– testes or wild-type ovaries. Two of the SSDS browser tracks are reproduced from Fig. 5G. Enrichment values are defined as coverage across the indicated coordinates relative to mean coverage for chr9 (see Methods for details). Note that ovary sample O1 and the Ankrd31–/– adult sample are known to have poorer signal:noise ratios than the other samples (30, 43). (B) SSDS maps for the mo-2 region on chr13. The red arrow indicates an ANKRD31- dependent, PRDM9-independent peak. The dashed segment indicates a gap in the mm10 genome assembly. Note: For all SSDS coverage tracks in this study, reads mapping to multiple locations are included after random assignment to one of their mapped positions. However, the same conclusions are reached about ANKRD31-dependence and PRDM9-independence of signal on chr9 and chr13 if only uniquely mapped reads are used (data not shown).

11 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. S9. RMMAI accumulation and high-level DSB formation on mo-2 regions in oocytes. (A) Examples of zygotene oocytes showing the colocalization between blobs of IHO1 and REC114, MEI4 and MEI1, ANKRD31 and mo-2 FISH signal. Scale bars = 2 µm. (B) Substantially less DSB formation occurs in oocytes near the mo-2 region on chr9. SSDS signal is from (43). The X-PAR is shown for comparison [previously shown to be essentially devoid of DSBs in ovary samples (43)]. See fig. S8A for quantification.

12 bioRxiv preprint doi: https://doi.org/10.1101/536136; this version posted January 31, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fig. S10. Summary of PAR ultrastructure and molecular determinants of axis remodeling and DSB formation. Schematic representation of the meiotic Y chromosome loop/axis strucuture before X–Y pairing at the transition between zygonema and pachynema. The chromosome axis comprises the meiosis-specific axial proteins SYCP2, SYCP3, HORMAD1, and HORMAD2; cohesin subunits (only REC8 is represented); and the RMMAI proteins. On the non-PAR portion of the Y chromosome axis (left), RMMAI protein loading and DSB formation are partly dependent on HORMAD1 and ANKRD31, and strictly dependent on MEI4, REC114 (24), IHO1 (31), and presumably MEI1 (29). The DNA is organized into large loops, with a low number of axis-associated RMMAI foci. By contrast, in the PAR (right), the hyper-accumulation of RMMAI proteins at mo-2 minisatellites (possibly spreading into adjacent chromatin) promotes the elongation and subsequent splitting of the PAR sister chromatid axes. Short mo-2-containing chromatin loops stretch along this extended PAR axis, increasing the physical distance between the PAR boundary and the distal PAR sequences, including the telomere. The degree of RMMAI protein loading, PAR axis differentiation, and DSB formation are proportional to the mo-2 FISH signal (which we interpret as reflecting mo-2 copy number), and depend on MEI4 and more specifically ANKRD31.

13